CN111801629A

CN111801629A - Method for determining oscillations occurring in a measurement signal

Info

Publication number: CN111801629A
Application number: CN201980017797.7A
Authority: CN
Inventors: J·高尔; V·施万伯格; T·维尔芬德; A·博特
Original assignee: Robert Bosch GmbH
Current assignee: Robert Bosch GmbH
Priority date: 2018-03-09
Filing date: 2019-03-07
Publication date: 2020-10-20
Anticipated expiration: 2039-03-07
Also published as: EP3762790B1; CN111801629B; US20200409318A1; EP3762790A1; DE102018203574A1; WO2019170789A1; US11774921B2

Abstract

The invention relates to a method for determining oscillations occurring in a measurement signal (10; u, u _ s, y), comprising the following steps: receiving a measurement signal (10; u, u _ s, y); determining an extremum of the received measurement signal (10; u, u _ s, y); and the closed loop of the measurement signal (10; u, u _ s, y) is determined by: a) identifying a location of the measurement signal (10; u, u _ s, y), wherein one closed loop is formed by two half loops (15, 16) with the same oscillation width and opposite directions, b) storing the identified closed loop, c) removing the identified closed loop from the measurement signal (10; u, u _ s, y), and d) repeating steps a) to c) until a complete closed loop has been determined.

Description

Method for determining oscillations occurring in a measurement signal

Technical Field

The invention relates to a method for determining oscillations occurring in a measurement signal, which can be used, for example, in monitoring and/or adaptive control of technical systems such as heating, ventilation and air conditioning systems.

Background

So-called PID regulators for regulating, for example, heating, cooling and air-conditioning equipment, entertainment and domestic electronics or industrial processes, such as occur, for example, in the chemical industry, are frequently used because of their well-known and simple application. A PID regulator is a regulator having a proportional component, an integral component and a derivative component.

The parameters of such PID controllers are often set using Ziegler and nicols heuristics which, although giving good interference behavior especially in the case of severely delayed stabilization processes, give oscillatory guiding behavior (fuhrungsverhalten). That is, the ziegler and nicols processes are mainly suitable for processes whose load characteristics do not change significantly with time and in which excessive vibration does not cause interference.

Thus, if larger changes can be expected in process dynamics or in load characteristics, the regulator parameters are often set adaptively. For this purpose, the control variable is continuously monitored with respect to oscillations around the setpoint value. Once oscillations are ascertained, the regulator parameters are adapted such that oscillations in the regulating variable are eliminated ("Patents, Software and hardware for PID Control: an overview and analysis of the current art", IEEE Control Systems Magazine, 26(1), pages 42 to 54, 2006, by Li et al).

The quality and accuracy of the determination of the oscillations in the control variable are decisive for the control behavior, in particular for the guidance behavior that can be achieved with an adaptive controller.

In principle, the manipulated variable can be decomposed into oscillation components of the manipulated variable by means of a Fast Fourier Transformation (FFT). However, in the case of poor resolution in the time domain, a wide window here leads to good resolution in the frequency domain, which makes it difficult to detect the frequency of the change. The opposite effect is obtained in the case of a narrow window. The fast fourier transform assumption involves a periodic signal such that information about the exact point in time at which the oscillation starts or stops is lost.

From WO 2010/054657 a1 a control method is known in which the decay rate of the oscillations in the control parameters (stillgr |) of a PID controller is determined and the control parameters are increased as a function of the decay rate. A high damping rate indicates a highly oscillating system close to the stability limit, so that the gain of the control variable is reduced, while a low damping rate indicates an excessive damping, so that the gain of the control variable is increased.

Disclosure of Invention

According to a first aspect, the invention provides a method for determining oscillations occurring in a measurement signal having the features of claim 1. The method comprises the following steps: receiving a measurement signal; determining an extremum of the received measurement signal; and a closed loop of the measurement signal is determined. Here, the (overall) closed loop of the measurement signal is determined by: a) identifying (first) closed loops in the measurement signal, wherein one closed loop is formed by two half loops with the same oscillation width and opposite directions; b) storing the identified closed loops and c) removing the identified closed loops from the measurement signal, wherein the mentioned steps a) to c) are repeated until all closed loops have been determined. These closed loops correspond to oscillations occurring in the measurement signal.

The measurement signal is preferably a sensor signal of a sensor, as is used in a control loop, for example a control loop with a heating, ventilation and air conditioning system. However, the expression "measurement signal" also includes signals derived from the sensor signal, for example by estimation methods.

The measures according to the independent claims have the following advantages: even oscillations in the measurement signals of complex, real systems can be determined simply, quickly and robustly with different, nested and interacting settings of often different suppliers. In this case, these oscillations can be evaluated independently of the source of the measurement signal, in particular independently of the respective sensor type.

Preferably, for determining the closed loop of the measurement signal, a so-called rain flow (Rainflow) counting method is used. Rain flow counting methods originate from the field of materials mechanics and are used in the field of materials mechanics to determine the fatigue life of components. The name "rain flow counting" derives from the analogy of the flow of rain water on the top of a tower. The time axis of the relevant measurement signal is assumed to be rotated by 90 ° clockwise. In this case, the area of the measurement signal on which "rain" drips from one "roof" to the next, lower roof is determined and counted as a half ring (also referred to as half hysteresis or half cycle). In particular, the following regions are counted as half rings: i) (signal) region starting at one extremum and ending opposite the larger extremum; ii) a region starting at one extreme and ending at a location where rain falling from the upper extreme occurs; and iii) a region starting at one extremum and ending at the end of the time series. Two half-rings of the same size, that is to say of the same oscillation width extending in opposite directions, that is to say axisymmetrically about an axis perpendicular to the time axis, constitute a closed loop (also called closed hysteresis or complete cycle). Extrema consist of local minima and maxima and are also referred to as reversal points (see, e.g., "Z ä hlverfahren und latannahme in der betriebsfetitgkeit" (counting method and load assumption in terms of operating intensity) "of mth baby-zone et al, chapter 2.4.4, pages 23-27, DOI 10.1007/978-3-642-.

The rain flow counting method allows the amount of data required for determining the oscillations contained in the measurement signal to be reduced in a concise manner even in the case of measurement signals with a high data volume, so that an efficient and time and resource-saving analysis is possible, wherein detailed information about the oscillations occurring in the measurement signal can be obtained.

The oscillation width of a half-loop is thus the amplitude of that oscillation occurring in the measurement signal which corresponds to the closed loop formed by the half-loop. It is also possible to simply determine the average value of the oscillation as half the amplitude of the half-loop of the oscillation. Furthermore, the period duration of the oscillation occurring in the measurement signal is derived from the double duration of the half-ring assigned to this oscillation. Furthermore, the starting time of the oscillation occurring in the measurement signal corresponds to the starting time of the complete loop assigned to the oscillation, i.e. the starting time of the first half loop of the complete loop in time. The corresponding situation applies to the end point in time of the oscillation. The time analysis enables a historical analysis of the appearance and disappearance of oscillations in the measurement signal. Advantageously, no prior information about the oscillations to be expected in the measurement signal is required here.

Preferably, limit values with respect to the oscillation width/amplitude of the half-rings and their duration (and thus the period duration of the oscillations) can be predefined in order to separate the oscillations to be considered from the oscillations not to be considered. For example, in the case of temperature measurement: half-width/oscillation with an amplitude of less than 1 ℃ and oscillation with a cycle duration of more than 1 hour are disregarded.

With the above-described method for determining oscillations occurring in measurement signals, it is possible to systematically check the measurement signals of a technical system, wherein the measurement data of heating, ventilation and air conditioning installations, for example of the same type but in different locations, for example installed in different buildings, can be compared with one another. This can be done, for example, by means of a central network platform.

In a method for monitoring a technical system, in particular a heating, ventilation and air conditioning system, which uses the above-described method for determining oscillations occurring in a measurement signal, an alarm can be output to a user or to a service person or to a user group if, for example, the amplitude (oscillation width) and/or the frequency of the determined oscillations occurring in the measurement signal exceeds a predefined limit value. The output of the alarm can be performed by means of a display and/or an alarm tone and/or by email and/or by electronic messages according to other protocols. In a preferred embodiment of the invention, the output or notification of the alarm can be assigned to different recipients depending on the magnitude or the sum of the magnitudes. This has the following advantages: at higher deviations, the corresponding technical fault may deteriorate and may be reported to a larger group of users, so that corresponding measures may be introduced.

It is also possible to provide a counter which is incremented, for example by 1, each time a closed loop, that is to say an oscillation, is determined. If the counter and thus the number of oscillations exceeds a predetermined limit value, an alarm can be output. Furthermore, the amplitudes of the oscillations occurring within a certain period of time, for example 1 hour, can be added and an alarm can only be output if the sum of the added amplitudes exceeds a predetermined limit value. Fault diagnosis may be performed after monitoring of the system. Furthermore, the determined signal (from within the determined closed loop) may also be used as an input to a data mining (Datamining) algorithm for fault searching and fault isolation or more accurate fault localization. Thus, for example, production faults can be detected when they occur more frequently from a certain magnitude. The determined signals can also be used for better fault localization, in particular in complex, multi-component technical systems or installations, in that it is possible to check in which subassembly of the device the most intense oscillations occur.

In addition to or instead of monitoring, the above-described method for determining oscillations occurring in the measurement signal can be used for adaptive control of technical systems, in particular heating, ventilation and air conditioning systems. In the case of adaptive regulation, the regulator parameters are adapted to the system and to changes in the process dynamics. Adaptive regulation is also referred to as self-learning regulation.

In this case, the parameters of the adaptive regulator are set as a function of the oscillations occurring in the respective measurement signal. Thus, the regulator parameters are preferably set such that the amplitude of the oscillation falls below a predetermined minimum height. Within the framework of the method for determining oscillations occurring in the measurement signal, the method for calculating the regulator parameters can be based on a minimum error square method, for example. Preferably, a PID regulator is used as the adaptive regulator. However, other regulator types, such as state regulators, may also be used alternatively or in combination. The regulator may be a multi-dimensional regulator. The above-described method for determining oscillations occurring in a measurement signal therefore provides a cross-system, systematic approach for setting parameters for adaptive adjustment.

Through the proposed monitoring and setting of the adaptive control, in particular a wrongly configured control can be identified and avoided, unusual behavior of the system components, for example, which can lead to a reduction of their service life, is ascertained and a proactive maintenance is introduced.

According to a further aspect of the invention, a computing unit is provided, which is set up to carry out the above-described method for determining oscillations occurring in a measurement signal, the above-described method for monitoring an adaptation and/or the above-described method for setting an adaptation. A computer program is also provided, which is set up to carry out the above-described method for determining oscillations occurring in a measurement signal, the above-described method for monitoring an adaptation and/or the above-described method for setting an adaptation. The computer program may be stored on a machine readable storage medium, such as an EEPROM, a flash memory or a data carrier, such as a DVD. Provision may also be made for: the computer program is provided on a central server, which can be implemented on the central server or downloaded from the central server via a computer network (internet, intranet, etc.). Access to the computer program may be through a network platform.

Further advantageous embodiments of the invention emerge from the following description and the accompanying drawings.

Drawings

Fig. 1 shows a flow chart of a method for determining oscillations occurring in a measurement signal according to an embodiment of the invention;

FIG. 2 shows a schematic diagram illustrating the determination of a closed loop in an exemplary measurement signal;

FIG. 3 shows a schematic diagram illustrating the removal of the determined closed loop from the measurement signal;

FIG. 4 illustrates an exemplary graph of monitoring of a compressor;

FIG. 5 shows a block diagram of a one-dimensional regulation loop with an adaptive regulator;

FIG. 6 shows a block diagram of a multidimensional regulation loop with adaptive regulation; and

fig. 7 shows a schematic view of an embodiment in which the invention is used to monitor a plurality of locations.

Detailed Description

Fig. 1 shows a flow chart of a method for determining oscillations occurring in a measurement signal. The measurement signal may be, for example, the volume flow, the pressure upstream or downstream of the compressor, or the temperature.

In step 1, the measurement signal is received by a calculation unit, for example a computer or an electronic control unit (not shown). In a preferred, optional step 2, the received measurement signal is checked for so-called outliers and these outliers are removed from the measurement signal. A measured value is classified as an outlier if it deviates from its previous measured value and/or its subsequent measured value by a specific, predefined value.

If the measurement signal is an analog signal instead of a digital signal, the measurement signal is preferably quantized in step 3. The order of steps 2 and 3 may be interchanged.

In a next step 4, an extremum of the measurement signal is determined, for example by calculating first and second derivatives of the measurement signal. In order to determine the extreme values, tolerances can be predefined which predetermine a minimum value by which these extreme values should deviate from one another, in order, for example, to ignore the effects of minimum oscillations and noise and thus to increase the accuracy of the oscillation determination.

In a next step 5, it is checked, preferably by means of the rain flow counting method described at the outset, whether the measurement signal is closed-loop (complete cycle, complete hysteresis). Here, the half-loops (half-cycles, half-hysteresis) present in the measurement signal are determined. A closed loop is obtained by two half-loops with the same oscillation width but in opposite directions.

If it is ascertained in step 6 that the measurement signal does not contain a single closed loop, the method ends in step 7. No oscillation can be determined in the measurement signal.

If, on the other hand, in step 6 it is ascertained that a closed loop and thus an oscillation can be determined, then this closed loop and thus this oscillation are stored in step 8, preferably together with their amplitude, starting point in time, ending point in time, period duration/frequency and/or mean value, in a storage medium (not shown), for example a RAM, and the closed loop found is removed from the measurement signal, the remaining parts of the measurement signal being connected to one another.

Fig. 2 is used to illustrate step 5 of fig. 1. The measurement signal, which is illustrated schematically and schematically in fig. 2 and relates to the temperature T, for example, has four

extreme values

11, 12, 13, 14 at time points T11, T12, T13 or T14 in the time domain shown. For a first half-ring 15 whose oscillation width extends from a maximum value 12 to a minimum value 13 and which is found by means of the rain flow counting method, a corresponding second half-ring 16 can be found in the measurement signal 10 by means of the rain flow counting method, which has the same oscillation width in the opposite direction. A first half-loop 15 is found, since it starts at a maximum value 12 and ends at a minimum value 13, which is opposite to the next maximum value 14 in time, which maximum value 14 has a greater value than the (starting) maximum value 12 (condition i of the rain flow counting method mentioned at the outset). The second half-ring 16 is found because it starts at the extreme value (minimum value 13) and ends at the location 17 where "rain" occurs, which "rain" drips "from the maximum value 12 in the direction of the time axis" downwards "(condition ii) of the rain flow counting method mentioned at the beginning). The first half-ring 15 and the second half-ring 16 together form a closed loop, which corresponds to an oscillation in the measurement signal 10.

In step 8 of fig. 1, the closed loop identified, that is to say the corresponding oscillation and its characteristic, is stored and deleted from the measurement signal 10, the remaining parts of the measurement signal being connected to one another. The latter is illustrated in fig. 3. That is to say the dashed triangle 18 in fig. 2 is removed and the measurement signal shown in fig. 3 is obtained, which is then searched for further closed loops by means of the rain flow counting method.

Thus, by means of the method illustrated by way of example in fig. 1, oscillations occurring in the measurement signal can be determined. Fig. 4 shows, for an example of a compressor of a heating, ventilation and air conditioning system, the oscillations per hour of the day determined by means of the method in a measurement signal, which is, for example, the pressure downstream of the compressor. Here, the amplitudes of the oscillations determined for the respective hours are added and averaged over the entire year. As can be seen, there is a drop in the height of the added amplitudes between 7:00 o 'clock and 20:00 o' clock. From this signal (course of change) it can be concluded that: outside the above-mentioned time periods corresponding to the main operating phases, the compressor must be clocked more frequently with lower load requirements. By this means, it can be automatically determined that higher losses occur outside the main operating phase.

Fig. 5 shows a schematically illustrated control circuit 20, which has: an adaptive regulator 21, which is preferably configured as an adaptive PID regulator; a control link (Stellglied) 22, such as a valve; and an adjustment object (Regelstrecke) 23, which is provided by the technical system 23 to be adjusted, for example, a heating, ventilation and air conditioning system. The technical system 23 of fig. 5 is a simple, i.e., one-dimensional system which has only one input signal (output variable u _ s of the control element) and only one output signal (control variable y).

The output signal (control variable u) of the adaptive controller 21, the output variable u _ s of the control element and the control variable y are supplied as measurement signals to the functional block 24. In a functional block 24, the method shown in fig. 1 for determining the oscillations occurring in the respective measurement signal is carried out on each of these measurement signals. The oscillations determined in the function block 24 or their characteristics, such as amplitude, cycle duration, etc., are supplied to the function block 25 for monitoring and fault diagnosis of the technical system 23, so that, for example, in the case of oscillations of high amplitude, the maintenance period can be shortened. The oscillations determined in block 24 are also used to recalculate and set the parameters of the adaptive regulator 21 in block 26 such that the oscillations of the measurement signals u, u _ s and y are minimized. The

blocks

21, 24, 25, 26 may be implemented separately or collectively by software and/or hardware, e.g. by a computer program running on a computing unit, e.g. a server.

Fig. 6 is an extension to the multidimensional regulation loop 30 with a complex technical system 33 with multiple input signals and output signals, as compared to fig. 5. In the case of local regulation shown in fig. 5, the parameters of the adaptator 21 can be set appropriately by means of the function block 26 only if the variables/signals in which oscillations occur in the event of a misconfiguration of the adaptator 21 are also input signals of the function block 24 responsible for the adaptator 21.

For the multidimensional control loop 30 shown in fig. 6, a central measurement signal detection and automation unit 37 is proposed, which detects all relevant measurement signals. The measurement signal detected by the measurement signal detection and automation unit 37 is supplied to a functional block 34, which implements the method shown in fig. 1 for determining the oscillations occurring in the measurement signal. The oscillations determined in the function block 34 or their characteristics are supplied to a function block 35 for monitoring and fault diagnosis of the technical system 33. In addition, the oscillations determined in the function block 34 and their characteristics are supplied to a function block 36 for adapting the regulator parameters. In

block

36 or 37, the correlation between the identified oscillations and the adaptive regulator used in the control loop 30 is preferably determined, and the regulator parameters of the regulators most strongly correlated with the respective oscillations are adapted such that the oscillations associated with these regulators are minimized. The adjustment is performed by a central measurement signal detection and automation unit 37. The functional blocks 34, 35, 36, 37 can be implemented jointly or separately in the form of software and/or hardware, in particular as computer programs on a server.

Advantageously, the method according to the invention for determining oscillations occurring in a measurement signal can be used centrally for parallel monitoring and/or regulation, in particular of a plurality of locations 43 (that is to say a plurality of technical systems), such as a plurality of production facilities or buildings (see fig. 7). Thus, for example, heating, ventilation and air conditioning systems of different buildings can be monitored and/or regulated centrally in parallel. All detected measurement signals of the position 43 are supplied to a central functional block 44, which determines the oscillations occurring in these measurement signals and their characteristics by means of the method shown in fig. 1. The function block 44 is preferably implemented as software (computer program) running on a central server. The server and the determined oscillations are now accessible via the network platform 45, and can be centrally analyzed and displayed on the network platform for monitoring and fault diagnosis purposes. To this end, corresponding monitoring and diagnostic software installed on the server may be accessed through the network platform. By means of the network platform, an adaptive adaptation of the regulator parameters can also be brought about, wherein the recalculation of the regulator parameters can likewise be effected by means of software installed on the server. In this way, for example, cross-location performance analysis and, if necessary, energy consumption reduction can be performed.

Claims

1. Method for determining oscillations occurring in a measurement signal, having the following steps:

-receiving a measurement signal (10; u, u _ s, y);

-determining extrema (11, 12, 13, 14) of the received measurement signal (10);

-determining a closed loop of the measurement signal (10; u, u _ s, y) by:

a) identifying a location of the measurement signal (10; u, u _ s, y), one of which is formed by two half-rings (15, 16) of the same oscillation width and opposite directions,

b) the identified closed loop is stored and,

c) removing the identified closed loop from the measurement signal (10), and

d) repeating steps a) to c) until a complete closed loop has been determined.

2. Method according to claim 1, wherein a rain flow counting method is used for determining the closed loop of the measurement signal (10; u, u _ s, y).

3. The method according to claim 1 or 2, wherein the amplitude of the oscillation occurring in the measurement signal (10; u, u _ s, y) is determined as a function of the oscillation width of the half-rings (15, 16) of the closed loop corresponding to the oscillation; the period duration of the oscillations occurring in the measurement signal (10; u, u _ s, y) is determined from the double duration of the half-rings (15, 16); and/or the starting point in time of the oscillation occurring in the measurement signal (10; u, u _ s, y) is determined as a function of the starting point in time of the half-rings (15, 16).

4. The method of any preceding claim, wherein a counter is incremented for each determined closed loop.

5. The method according to any of the preceding claims, wherein an outlier is removed from the measurement signal (10; u, u _ s, y) before determining an extremum (11, 12, 13, 14) of the received measurement signal (10; u, u _ s, y).

6. Method for monitoring a technical system (23; 33; 43), in particular a heating, ventilation and air conditioning system, comprising the following steps:

-determining, with the method according to any of the preceding claims, for one or more measurement signals (10; u, u _ s, y), oscillations occurring in the respective measurement signal (10; u, u _ s, y);

-outputting an alarm if the amplitude of the oscillations occurring in the measurement signal (10; u, u _ s, y), the number of oscillations occurring in the measurement signal (10; u, u _ s, y) and/or the added amplitude of the oscillations occurring in the measurement signal (10; u, u _ s, y) exceeds a predetermined limit value.

7. The method according to claim 6, used for the parallel monitoring of a plurality of technical systems (43), wherein the method can be implemented by a network platform (45).

8. Method for the adaptive control of technical systems (23; 33; 43), in particular heating, ventilation and air conditioning systems, comprising the following steps:

-determining, with the method according to any one of claims 1 to 5, oscillations occurring in one or more measurement signals (10; u, u _ s, y) for the respective measurement signal (10; u, u _ s, y);

-setting parameters for an adaptive regulator (21), in particular an adaptive PID regulator, of the technical system (23; 33; 43) as a function of oscillations occurring in the respective measurement signal (10; u, u _ s, y).

9. The method of claim 8, wherein the technical system (33) has a plurality of adaptors and the parameters of each adaptor are set in accordance with those oscillations associated with the respective adaptor.

10. The method according to claim 8 or 9, used for adaptive tuning of a plurality of technical systems (43), wherein the method can be implemented by a network platform (45).

11. A computing unit, which is set up to carry out the method according to one of the preceding claims.

12. Computer program which, when implemented on a computing unit, causes the computing unit to carry out the method according to any one of claims 1 to 10.

13. Machine readable storage medium having stored thereon a computer program according to claim 12.